Muhammad al-xorazmiy nomidagi toshkent axborot texnologiyalari universiteti kompyuter Injiniring Fakulteti

Figure 1. Search strategy Preferred Reported Item for Systematic review and Meta-Analysis (PRISMA) flow chart.

3.2. Risk Assessment

Twenty-one and 38 studies were assessed as being prone to medium and low risks of bias, respectively (Table 2). None of the considered articles scored lower that six on the employed appraisal checklist. All reviewed articles presented reliable measurements (Q5) and conclusions that were justified by their results (Q10); yet, many authors have inadequately reported or justified sample characteristics (Q3, 37%), study limitations (Q11, 53%) and funding or possible conflict sources (Q12, 51%). Statistics (Q6, 81%) and general methods (Q7, 88%) were typically described in depth. Generally, studies were favourably assessed against all the remaining items of the employed appraisal tool (Q1, 95%; Q2, 92%; Q4, 93%; Q8, 98%; Q9, 93%). The assessments of both reviewers were consistent and comparable with average review scores of 9.9 ± 1.6 and 9.9 ± 0.9.
Table 2. Risk of bias assessment.

3.3. MoCap Technologies in Industry

In the reviewed studies, pose and position estimation was carried out with either inertial or camera-based sensors (i.e., RGB, infrared, depth or optical cameras), or in combination with each other (Table 3). Inertial sensors have been widely employed across all industry sectors (49.2% of the reviewed works), whether the tracked object was an automated tool, the end effector of a robot [30,37,64], or the operator [27,36,39]. In 30.5% of the reviewed studies, camera-based off-the-shelf devices such as RGB, IR and depth cameras, mostly coming from the gaming industry (e.g., Microsoft Kinect and Xbox 360), were successfully employed for human activity tracking, and gesture or posture classification [25,77]. Inertial and camera-based sensors were used in synergy in 10.2% of the considered works, in the tracking of the operator’s body during labour or the operator’s interaction with an automated system (e.g., robotic arm). EMG, ultra-wide band (UWB) nets, resistive bending sensors or scanning sonars were used along with IMUs to improve pose and position estimation in five studies (8.5%). One study also coupled an IMU sensor with a CCTV and radio measurements. Generally, IMU and camera-based sensors were used consistently in the industry during the last 5 years (Figure 2).